Screening in digital monopolies

This paper analyzes quality-based screening in digital monopolies, demonstrating that costless replication leads to a dual inefficiency where the monopolist underinvests in the highest quality while distributing degraded versions, a dynamic where competition worsens underinvestment but improves distributional outcomes.

The Gist

Here is an explanation of the paper "Screening in digital monopolies" using simple language, analogies, and metaphors.

The Big Idea: The "Master Cake" vs. The "Car Factory"

Imagine you are a baker. You have two ways to make cakes:

1. The Car Factory Model (Old School): To make a basic cake, you need basic ingredients. To make a fancy cake with gold leaf, you need extra expensive ingredients. If you want to sell both, you have to pay for the gold leaf only for the fancy cakes. The cost of the basic cake doesn't depend on the fancy one. This is how most physical goods (like cars or clothes) work.
2. The Master Cake Model (Digital Goods): Imagine you bake one giant, perfect "Master Cake." It costs a lot of time and money to bake this one masterpiece. But once it's done, you can slice it up.
   • To sell a "Premium" slice, you give someone a big, perfect piece.
   • To sell a "Basic" slice, you just take a knife and cut off the fancy frosting or crumble the edge. You don't need new ingredients; you just damage the Master Cake.
   • The Twist: The cost of making the "Basic" slice is zero. The only cost you paid was for the Master Cake.

This paper is about what happens when a company (a monopolist) sells digital goods (like software, movies, or data) using the Master Cake model.

---

The Problem: The "Digital Damaging" Dilemma

In the old world (cars), if a company wants to sell cheap and expensive versions, they build them separately. In the digital world, the company builds the best version first, then intentionally makes "worse" versions by turning off features (like disabling 4K video or hiding a "Pro" button).

The authors ask: Does this change how companies behave?

1. The "Under-Baking" Mistake (Productive Inefficiency)

In a normal market, a company might build a great car and sell it to rich people, and a cheaper car to poor people. They build the best car because the rich people pay for it.

But in the Digital Master Cake world, the company knows that if they make the cake too perfect, they have to sell the "damaged" (basic) slices to everyone else at a lower price.

• The Result: The company decides to bake a smaller, less perfect Master Cake than what would be socially ideal. They intentionally hold back on quality to save money on the initial "baking" cost.
• Analogy: Imagine a movie studio. Instead of spending $200 million to make a blockbuster with the best special effects, they spend $100 million. Why? Because if they spend the extra $100 million, they have to lower the price of the "budget" streaming version for everyone else to keep their profits up. So, they under-invest in quality.

2. The "Damaged Goods" Mistake (Distributional Inefficiency)

Because the company made a smaller Master Cake, they have to be very clever about who gets what.

• The Rich (High Types): They get the full, undamaged cake.
• The Poor (Low Types): They get the crumbled, damaged slices.
• The Inefficiency: In a perfect world, everyone should get the same high-quality cake because it costs nothing to copy it! But the company damages the cake for the poor people just to trick them into thinking they are getting a different product, so the rich people will pay more.

---

What Happens When They Compete?

The paper also asks: What if two digital companies fight?

In a normal market, competition usually makes things better and cheaper. In this digital world, it's a mixed bag:

1. The "Free" Slice: Competition is so fierce that the second-best company is forced to give away their "Basic" slices for free.
2. The "Race to the Bottom": Because companies are terrified of losing the "Premium" customers to a rival, they get scared to invest in a huge Master Cake. They end up baking even smaller cakes than the monopoly did.
   • Result: The highest quality available in the market actually drops when there is competition.
3. The Good News: However, because the "Basic" slices are free and there are more companies, more people get access to some quality. The "damaging" (holding back quality) happens less often.

Summary of Competition:

• Bad: The top quality gets worse (companies are too scared to invest).
• Good: More people get a decent product for free or cheap.

---

The "No-Damaging" Rule (Regulation)

What if the government says, "You cannot damage your product! You must sell the same quality to everyone"?

• The Result: The company is forced to sell the same quality to everyone.
• The Twist: Because they can't use "damaged" slices to trick the rich into paying more, they decide to bake an even smaller Master Cake than before.
• Why? Without the ability to "damage" the product to separate customers, the company feels it's not worth investing in high quality at all. They might even stop selling to the poorest customers entirely.
• Lesson: Sometimes, banning "bad versions" of software might actually make the "good version" worse for everyone.

---

The Takeaway in One Sentence

In the digital world, because it's cheap to copy and easy to break things, companies often under-invest in quality and intentionally break their products to sell them at different prices; while competition helps more people get access, it can actually make the top-tier quality even worse.

The "Digital Economy" Metaphor

Think of the digital economy like a giant library.

• The Old Way: You have to print a new book for every reader. If you want a hardcover and a paperback, you pay for paper and ink for both.
• The Digital Way: You write one perfect book. You can photocopy it a million times for free. But to sell it to different people, you have to tear out the last 50 pages for the "Budget" edition.
• The Paper's Conclusion: Because tearing pages is free, the author might decide to write a shorter book in the first place so they don't have to worry about the "Budget" edition being too good. And if two authors compete, they might write even shorter books just to avoid the risk of being undercut!

Technical Summary

Here is a detailed technical summary of the paper "Screening in digital monopolies" by Pietro Dall'Ara and Elia Sartori.

1. Problem Statement

The paper investigates quality-based screening (versioning) in markets for digital goods, where the cost structure fundamentally differs from traditional physical goods.

• The Core Feature: Digital goods exhibit costless replication and degradation. Once a high-quality "cap" version is produced, the seller can create and distribute lower-quality versions (by "damaging" or restricting features) at zero marginal cost.
• The Departure from Standard Theory: Standard screening models (e.g., Mussa-Rosen) assume cost separability, where the total cost is the sum of costs for each individual unit produced. In digital markets, the total production cost depends only on the highest quality supplied (the "cap"), not on the number of units or the specific qualities of lower tiers.
• Research Questions: How does this non-separable cost structure alter the monopolist's optimal allocation? How does it interact with information rents? What are the welfare implications of competition and regulatory bans on versioning in such markets?

2. Methodology

The authors develop a mechanism design model with the following specifications:

• Agents: A monopolist seller and a continuum of buyers with types $\theta \in [0, 1]$ distributed according to $F$.
• Utility: Buyers have utility $u(q, \theta) = g(q) + \theta q$, where $g(q)$ is a concave, increasing function representing common utility from quality, and $\theta q$ represents type-dependent marginal utility.
• Cost Structure: The cost of producing and distributing an allocation $q(\cdot)$ is $c(\sup_{\theta} q(\theta))$. The cost function $c$ is strictly convex and increasing, depending solely on the maximum quality (the cap) provided.
• Decomposition Approach: To solve the non-separable problem, the authors decompose the monopolist's problem into a family of separable sub-problems.
  1. The Cap-Constrained Problem ($P(q)$): For a fixed cap quality $q$, the seller solves a standard virtual-surplus maximization problem subject to the constraint that no buyer receives quality higher than $q$.
  2. The Acquisition Problem: The seller chooses the optimal cap $q^M$ by equating the marginal cost of increasing the cap, $c'(q)$, with the marginal revenue generated by relaxing the cap constraint.

3. Key Contributions and Theoretical Results

A. The Monopolist Allocation

The optimal allocation $q^M(\theta)$ takes the form of a truncated virtual-surplus maximizer:
$$q^M(\theta) = \min\{\beta(\theta), q^M\}$$
where $\beta(\theta)$ is the quality that maximizes virtual surplus in a zero-cost environment, and $q^M$ is the chosen cap.

• Two Interdependent Inefficiencies:
  1. Productive Inefficiency: The monopolist underinvests in the highest quality. The chosen cap $q^M$ is strictly lower than the socially efficient quality $q^*$. This occurs because the marginal revenue of increasing the cap is lower than the marginal social value (due to the "bunching" of high types).
  2. Distributional Inefficiency: Buyers below a "marginally bunched" type $b(q^M)$ receive a damaged (lower) quality than they would under the unconstrained virtual-surplus maximizer.
• The Marginally Bunched Type ($b(q)$): This is a critical threshold type.
  • Types $\theta > b(q^M)$ receive the undamaged cap quality $q^M$.
  • Types $\theta < b(q^M)$ receive a damaged quality determined by the standard screening logic (virtual surplus maximization) but capped at $q^M$.
  • The marginal revenue from increasing the cap comes entirely from the mass of types above $b(q^M)$, weighted by the marginal utility of type $b(q^M)$.

B. Comparative Statics

• Cost Sensitivity: An increase in production costs (scaling $c$) reduces the cap $q^M$ and expands the bunching region (more types receive the cap).
• Utility Curvature: An increase in the curvature of the common utility component $g$ (parameter $\kappa_g$) increases the cap and expands the bunching region. Rich versioning (screening) requires intermediate curvature; if curvature is too low (linear), the solution collapses to exclusion or pooling.

C. No-Screening Constraint (Regulatory Ban)

If the seller is banned from damaging goods (must offer a single quality or exclude):

• Result: The productive inefficiency worsens. The monopolist chooses an even lower cap ($q^M_N < q^M$) because the marginal revenue of increasing the cap is lower without the ability to extract rents via damage.
• Distributional Effect: A larger share of buyers receives the undamaged good (distributional efficiency improves), but low types may be excluded entirely, making them worse off.

D. Competition

The authors model competition as a two-stage game: firms commit to a quality cap, then compete in prices (Bertrand).

• Equilibrium: In pure strategies, a monopoly emerges. In mixed strategies (with $n \ge 2$ active firms), firms randomize over caps.
• The "War of Attrition": Competition creates a war-of-attrition dynamic in the production stage. The second-highest quality in the market is offered for free (Bertrand competition drives the price of the second-highest tier to zero).
• Welfare Implications:
  • Productive Inefficiency: Competition exacerbates underinvestment. The highest quality produced in equilibrium is strictly lower than the monopoly cap ($X < q^M$).
  • Distributional Efficiency: Competition improves distribution. More types receive undamaged goods (the second-highest quality is free and available to low types).
  • Net Welfare: The effect on total welfare is ambiguous and depends on the shape of the cost function. If costs are steep (fixed-cost-like), monopoly may dominate due to higher quality investment. If costs are flat, competition may dominate due to better access.

4. Significance and Implications

• Digital Economy Relevance: The paper provides the first rigorous mechanism design framework specifically for digital goods where replication and degradation are costless. It explains why digital monopolists (e.g., software, streaming, data) often underinvest in top-tier features relative to social optimality while offering a wide array of "damaged" versions.
• Policy Insights:
  • Regulation: Bans on versioning (forcing a single quality) may backfire by reducing the quality of the top tier, hurting high-value users without necessarily helping low-value users (who might be excluded).
  • Antitrust: Competition in digital markets does not necessarily lead to higher quality. While it improves access (more free tiers), it can lead to a "race to the bottom" in terms of the maximum quality produced due to the erosion of profit margins needed to fund R&D (the "cap").
• Theoretical Advancement: The paper extends the Mussa-Rosen framework by introducing production externalities (where the cost of serving one type depends on the quality assigned to others). It demonstrates that the "efficiency at the top" result of standard screening models does not hold; instead, the top type suffers from productive inefficiency because their quality level dictates the cost for the entire population.

Summary of Inefficiencies

Feature	Standard Separable Costs (Mussa-Rosen)	Digital Non-Separable Costs (This Paper)
Top Type Quality	Efficient (no distortion at the top)	Productively Inefficient (Underinvested)
Low Types	Distorted downward (damaged)	Distorted downward (damaged)
Source of Inefficiency	Information rents only	Interaction of information rents + Production Externality
Effect of Competition	Generally improves welfare	Ambiguous: Improves access but reduces max quality

This paper fundamentally reshapes the understanding of monopoly power in digital markets, highlighting that the unique cost structure of digital goods creates a specific trade-off between the level of quality produced and the breadth of access, which is distinct from traditional physical goods markets.